When an FM signal propagates by more than one path, delayed replicas appear in the receive passband. The signals can interfere with each other and distort the recovered audio. Multipath distortion is nonharmonic, nonmusical, and can be very annoying. Rotating an outdoor antenna or repositioning an indoor one can lower the distortion by reducing antenna response to secondary paths. But the character and audibility of the distortion depend strongly on the program material. Low levels of it can be difficult to identify, making antenna aiming tedious and ambiguous.
Some FM tuners indicate the level of multipath interference on a meter. Others let you press a button to hear a special audio signal affected by multipath. But these methods often don't work well. They suppress or alter the distortion character and hide its subjective impact. In addition, residual program material often masks the multipath signal, particularly when using narrow IF filters. Oscilloscope multipath indicators can be more revealing, but they also can be difficult to interpret, particularly when inaudible co-channel or adjacent-channel signals are present. These methods all use the envelope of the IF signal.
Multipath distortion mainly corrupts the L−R stereo subchannel, which extends from 23 to 53 kHz. L−R information is transmitted on an amplitude-modulated subcarrier at 38 kHz. The carrier is suppressed and replaced by a 19 kHz pilot signal. The stereo decoder locks a 38 kHz local oscillator to the pilot and uses it to synchronously demodulate the L−R signal. If the oscillator is not in phase with the pilot, detected L−R amplitude drops. When the oscillator is phased exactly 90°, no signal is recovered.
This suggests a way to provide an audible window into the stereo subchannel region that avoids program material. Simply demodulate the subcarrier region with a quadrature oscillator (one phased 90°). For a perfectly transmitted and received signal, you'll get no output. For a real signal you may hear L+R harmonics, phase-rotated L−R intermodulation products, or crossmodulation between L+R, L−R, SCA, RDS, or HD Radio sidebands. Multipath propagation can cause any of these artifacts. You may also hear co-channel interference, adjacent-channel interference, HD Radio self-noise, or background noise. I call the function that provides the diagnostic signal a Quadrature Multipath Monitor (QMM).
This QMM circuit is suitable for any tuner whose stereo decoder provides a 76 kHz VCO signal. I substitute the QMM output for the Cal Tone signal in my Sony ST-S555ES. When I press the Cal Tone button, I hear whatever is in the L−R region, but the L−R and L+R signals are suppressed. Alternatively, you can route the QMM output to an auxiliary input on your preamp. In addition to the audio path, you might connect the QMM signal to an oscilloscope so you can observe it without having to switch away from normal audio.
The circuit uses two sections of a 74HC4066 quad analog switch as a synchronous demodulator. A TL074 quad op-amp acts as a comparator, inverter, and buffer. The stereo decoder's 76 kHz sawtooth waveform triggers a J-K flip-flop that drives the switches. (The 74C76 ground pin is at −5V so do not use a 74HC76.) The 5kΩ pot sets the trigger phase such that the 74C76 38 kHz output is in quadrature with the L−R signal. Adjust this pot to minimize L−R feedthrough. Adjust the 20kΩ balance pot to minimize L+R feedthrough. Adjust the 100kΩ pot so that the QMM noise level matches that in normal stereo mode for a weak signal.
QMM provides best insight when it monitors the signal driving your stereo decoder. Connect it after any postdetection filter. The QMM input impedance is very high and it won't load a passive filter.
With QMM you may be able to find a propagation path that yields weaker but clearer signals by minimizing multipath distortion. I discovered that I had been somewhat mispointing my rotary antenna when aiming it toward Los Angeles from my location in northern San Diego County. My tuner's LED signal strength indicator peaked over a broad range, but with QMM I was able to find a particular direction that minimized multipath distortion.
After using QMM for quite some time I added multipath outputs to the ST-S555ES tuner and attached an oscilloscope. I found the visual multipath display much less revealing than QMM. It is possible to hear multipath effects with QMM without seeing anything unusual on the scope. Conversely, the scope responds to co-channel and adjacent-channel signals, often appearing to show multipath interference when none is present.
QMM can provide fascinating insight into FM signal quality. The level and character of extraneous sounds in the L−R region vary widely. Older analog transmit processors may reveal themselves. Harmonics of L+R sibilants may be heard, even on mono signals. Strange dissonances correlated with the L−R signal may appear. Because program material often dominates these sounds, QMM can help to isolate and identify sources of muddy audio.
QMM can expose tuner aberrations otherwise inaudible. Although I measure higher distortion on the bench, I've never been able to hear a difference in sound quality when switching between wide and narrow IF filters in any properly aligned tuner, except when multipath is present. But with QMM I can hear a difference. Similarly, on very strong signals I often measure higher audio distortion, perhaps due to RF AGC. Although I can't hear the effect on normal audio, the distortion products are audible with QMM. A way to align a tuner to minimize artifacts in the L−R subcarrier region audible with QMM is described here.
Detected noise in FM systems increases 6 dB per octave. After deemphasis, L−R noise is 23 dB stronger than L+R noise. Because QMM noise originates in the L−R region, QMM can function as an extra-sensitive background noise detector, even for monophonic signals. Use it for minimum-noise antenna aiming or other RF checks without the program material getting in the way.
This is the application circuit from the Sanyo LA3401 stereo decoder datasheet. C11 and the internal 1kΩ resistor at pin 3 form a phase-shift network that compensates for any differential IF delay between 19 and 38 kHz. Most stereo decoders use a similar scheme. If you increase C11 until the 19 kHz delay becomes 45°, the decoder will extract the quadrature L−R signal since the 38 kHz delay will be 90°. Ken Wetzel used .01 µF and added a 1kΩ trimpot in series with pin 3 for fine adjustment. To obtain the quadrature L−R signal alone, you must suppress the internal L+R injection. Adding 100 µF or more from pin 4 to ground will do this (Ken measured Ra as 5.5kΩ). Resistors Rc are external to the IC in some decoders. Breaking these paths is a more robust way to kill L+R injection. If you use separate switches to kill L+R and to shift phase as Ken did, you can listen to L−R alone. This may reveal other signal anomalies.
This sound sample illustrates what QMM can reveal. My antenna was aimed away from the station toward a distant mountain range illuminated by the signal: